fix(#6806): capture num_turns from incomplete Claude streams by fullsend-ai-coder[bot] · Pull Request #7687 · fullsend-ai/fullsend

fullsend-ai-coder · 2026-09-24T19:11:44Z

Summary

Claude Code review (and other) runs that crash or are killed after producing valid output no longer report num_turns: 0 in metrics.json. The runner now counts message_start events on the stream and carries that running total on TokensEvent, the same incremental channel already used for token counts.

Related Issue

This is the remaining gap from #6806 after #6908/#6938 captured token counters on cancelled/killed Claude runs. Those PRs left NumTurns and TotalCostUSD sourced only from the terminal ResultEvent.

Changes

Count message_start events in parseClaudeStream and attach the running turn count to TokensEvent.
Fold TokensEvent.NumTurns into RunMetrics via a shared applyClaudeMetrics helper (used by both ClaudeRuntime.Run and progressParser). A later ResultEvent still overwrites with authoritative totals.
Document that num_turns has this fallback in docs/cli/run.md.
Cost is unchanged. Claude Code only reports total_cost_usd on the terminal result line, and fullsend does not estimate cost from token counts (Cost data contract). Incomplete streams still persist total_cost_usd: 0.

This does not revive the synthetic-ResultEvent approach from closed PR #6808. Turns ride the existing TokensEvent snapshot, matching the #6908 architecture for tokens.

Testing

New regression: incomplete multi-turn stream reports NumTurns == 3 and TotalCostUSD == 0
Broken-pipe / cancelled-run tests assert non-zero turns
message_start without message_delta still emits a deferred TokensEvent with NumTurns
ResultEvent still overwrites the incremental snapshot
go test ./internal/runtime/ and targeted ./internal/cli/ aggregation tests pass
go vet ./internal/runtime/ ./internal/cli/
applyClaudeMetrics is 100% covered; parseClaudeStream is 93.2%

Checklist

PR title follows Conventional Commits (fix(#6806): ...)
Commits are not signed off (autonomous agent, DCO-exempt)
I wrote this contribution myself and can explain all changes in it

Closes #6806

Post-script verification

Branch is not main/master (agent/6806-incremental-num-turns)
Secret scan passed (gitleaks — fbf65c8b8c717d74fa6033519658851efaba91f9..HEAD)
PR body secret scan passed (gitleaks — no-git)

Claude Code only reports num_turns on the terminal result NDJSON line. When the process exits abnormally after doing work (exit_code -1), metrics.json recorded num_turns: 0 even though token counts already survive via TokensEvent (#6905). Count message_start events and carry the running turn count on TokensEvent so incomplete streams still populate num_turns. Dollar cost stays zero: the runtime does not report incremental cost, and fullsend does not estimate it from token counts. Note: pre-commit could not fetch remote hook repositories (HTTP 403). Local equivalents were run: gofmt, go vet, trailing-whitespace, end-of-file, lint-docs-links, lychee, lint-broken-symlinks, and scan-secrets. Closes #6806

fullsend-ai-review · 2026-09-24T19:13:35Z

🤖 Finished Review · ✅ Success · Started 7:13 PM UTC · Completed 7:27 PM UTC

Commit: 13c77fa · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $4.66

github-actions · 2026-09-24T19:15:37Z

Site preview

Preview: https://f9401bec-site.fullsend-ai.workers.dev

Commit: 7e03869e10d4b1db2a3dba4ed6fc9038788ba2df

codecov · 2026-09-24T19:17:29Z

Codecov Report

❌ Patch coverage is 96.96970% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/runtime/claude.go	0.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

fullsend-ai-review · 2026-09-24T19:26:57Z

Risk Assessment: moderate (2/5)

Details

Moderate risk: focused 7-file bug fix with strong test coverage by a bot author, offset by high churn and coupling in modified CLI and runtime files (unchanged from prior review; anchored).

Previous run

Risk Assessment: moderate (2/5)

Details

Moderate risk: focused 7-file bug fix with strong test coverage by a bot author, offset by high churn and coupling in modified CLI and runtime files.

Previous run (2)

Risk Assessment: moderate (2/5)

Details

Moderate risk: focused 7-file bug fix with strong test coverage by a bot author, offset by high churn and coupling in modified CLI and runtime files.

fullsend-ai-review · 2026-09-24T19:26:59Z

Looks good to me

Previous run

Looks good to me

Previous run (2)

Review

Findings

Low

[api-shape-pattern] internal/runtime/claude_progress.go:541 — applyClaudeMetrics dereferences metrics without a nil guard. Its counterpart applyCodexMetrics (internal/runtime/codex_progress.go:668-671) returns immediately on nil. This PR extracted the Claude fold into a package-level helper used from ClaudeRuntime.Run and progressParser; both still panic on a nil *RunMetrics. No current production or test call site passes nil (internal/cli/run.go uses var metrics ...; rt.Run(..., &metrics), and ClaudeRuntime{}.Run tests do the same), so this is a defensive consistency gap, not a live crash.
Remediation: Add if metrics == nil { return } at the start of applyClaudeMetrics to match applyCodexMetrics.

Next steps:

/fs-fix — agent addresses review findings automatically
/fs-fix <your instruction> — agent fixes with your specific guidance
Push commits directly — review re-runs automatically on push
/fs-fix-stop — disable automatic fix runs for this PR

fullsend-ai-coder · 2026-09-24T19:28:55Z

🤖 Finished Fix · ✅ Success · Started 7:28 PM UTC · Completed 7:38 PM UTC

Commit: 13c77fa · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $1.43

Add the same nil check applyCodexMetrics already has. No current call site passes nil, so this is a defensive consistency fix, not a live crash fix, per review feedback on PR #7687. Addresses #7687

fullsend-ai-coder · 2026-09-24T19:38:09Z

🔧 Fix agent — iteration 1 (bot-triggered)

Added the missing nil guard to applyClaudeMetrics to match applyCodexMetrics's defensive pattern, plus a regression test asserting it does not panic on a nil *RunMetrics.

Fixed (1):

applyClaudeMetrics dereferences metrics without a nil guard, unlike its counterpart applyCodexMetrics (internal/runtime/claude_progress.go): Added an if metrics == nil { return } guard at the top of applyClaudeMetrics, matching applyCodexMetrics. Verified no current call site (ClaudeRuntime.Run, progressParser) passes nil, so this is a defensive consistency fix rather than a live-crash fix, as the review itself noted. Added a regression line to the existing TestApplyClaudeMetricsTokensEventSetsTurns test asserting applyClaudeMetrics(nil, ResultEvent{}) does not panic, mirroring applyCodexMetrics's existing nil test.

Tests: passed

CI inspections (13):

test (passing, success): CI unit/integration test job passed on the pre-fix head SHA.
commit-lint (passing, success): Commit message linting passed.
web (passing, success): Web/docs build job passed.
test-sandbox-darwin (passing, success): Darwin sandbox test job passed.
codecov/patch (passing, success): Patch coverage check passed.
functional-tests (passing, success): Functional test suite passed.
gate (E2E Tests) (passing, success): E2E gate job passed.
gate (Functional Tests) (passing, success): Functional-tests gate job passed.
behaviour (passing, success): Behaviour test suite passed.
e2e (passing, success): E2E test suite passed.
build (passing, success): Docs site build job passed.
vouch-gate (passing, success): Contributor vouch gate passed.
DCO (passing, success): DCO check passed (autonomous commits exempt).

Next steps:

/fs-review — request a re-review of the changes
/fs-fix <your instruction> — run another fix pass with specific guidance
Push commits directly — review re-runs automatically on push
_{Updated by fullsend fix agent}

fullsend-ai-review · 2026-09-24T19:40:19Z

🤖 Finished Review · ✅ Success · Started 7:40 PM UTC · Completed 7:50 PM UTC

Commit: 3e04bd5 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $2.67

Superseded by updated review

fullsend-ai-review · 2026-09-27T18:49:10Z

🤖 Finished Review · ✅ Success · Started 6:49 PM UTC · Completed 7:02 PM UTC

Commit: 7e03869 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $2.81

Superseded by updated review

fullsend-ai-coder Bot requested a review from a team as a code owner September 24, 2026 19:11

fullsend-ai-coder Bot added the ready-for-review Triggers review agent dispatch label Sep 24, 2026

fullsend-ai-coder Bot temporarily deployed to dev September 24, 2026 19:12 Inactive

github-actions Bot deployed to site-preview September 24, 2026 19:15

View deployment

fullsend-ai-review Bot added the risk/moderate PR risk: moderate label Sep 24, 2026

This comment was marked as outdated.

Sign in to view

refactor(#6806): guard applyClaudeMetrics against a nil *RunMetrics

3e04bd5

Add the same nil check applyCodexMetrics already has. No current call site passes nil, so this is a defensive consistency fix, not a live crash fix, per review feedback on PR #7687. Addresses #7687

fullsend-ai-coder Bot had a problem deploying to dev September 24, 2026 19:38 Failure

github-actions Bot deployed to site-preview September 24, 2026 19:42

View deployment

This comment was marked as outdated.

Sign in to view

fullsend-ai-review Bot added the ready-for-merge All reviewers approved — ready to merge label Sep 24, 2026

Merge branch 'main' into agent/6806-incremental-num-turns

7e03869

ifireball temporarily deployed to dev September 27, 2026 18:47 — with GitHub Actions Inactive

github-actions Bot deployed to site-preview September 27, 2026 18:51

View deployment

fullsend-ai-review Bot approved these changes Sep 27, 2026

View reviewed changes

Conversation

fullsend-ai-coder Bot commented Sep 24, 2026

Summary

Related Issue

Changes

Testing

Checklist

Post-script verification

Uh oh!

fullsend-ai-review Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Site preview

Uh oh!

codecov Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

fullsend-ai-review Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

fullsend-ai-review Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review

Findings

Low

Uh oh!

This comment was marked as outdated.

Uh oh!

fullsend-ai-coder Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

fullsend-ai-coder Bot commented Sep 24, 2026

🔧 Fix agent — iteration 1 (bot-triggered)

Uh oh!

fullsend-ai-review Bot commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as outdated.

Uh oh!

fullsend-ai-review Bot commented Sep 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

fullsend-ai-review Bot commented Sep 24, 2026 •

edited

Loading

github-actions Bot commented Sep 24, 2026 •

edited

Loading

codecov Bot commented Sep 24, 2026 •

edited

Loading

fullsend-ai-review Bot commented Sep 24, 2026 •

edited

Loading

fullsend-ai-review Bot commented Sep 24, 2026 •

edited

Loading

fullsend-ai-coder Bot commented Sep 24, 2026 •

edited

Loading

fullsend-ai-review Bot commented Sep 24, 2026 •

edited

Loading

fullsend-ai-review Bot commented Sep 27, 2026 •

edited

Loading